Predict: Next gen console tech (9th iteration and 10th iteration edition) [2014 - 2017]

Status
Not open for further replies.
Now that they finally got consoles to x86, I doubt anyone in their right mind is even playing with the thought of going away from it again [in near future]

With Windows 10 running on ARM, I see a XboxTwo console on ARM more posible now than a year ago.
 
Windows 8 ran on ARM too, what's your point?

If I am not wrong, Windows RT was a special breed of Windows, and not "compatible". But Windows 10 is the same for all platforms.

I just saying that for me is more posible now that there is just one ecosystem for APPs, than a year ago.

My next gen console will be a MicroATX or MiniItX computer.
 
With Windows 10 running on ARM, I see a XboxTwo console on ARM more posible now than a year ago.
With AMD Zen in the horizon?. For first time in a long time consoles could have a great CPU in next gen... And more important, they will have the OS ready and 100% backwards compatibility from day one, apart from having game engines ready from day one too!.
 
If I am not wrong, Windows RT was a special breed of Windows, and not "compatible". But Windows 10 is the same for all platforms.
ARM devices won't run regular x86 windows software, only the "Universal Apps".
 
AMD released some juicy HBM information for fun and profit...
http://hothardware.com/news/amd-details-high-bandwidth-memory-hbm-interface-technology

One HBM stack is only 35mm2.
It's weird that they have 1GB per stack. Did they cheat the specs to have 8 Hi? Because it seems the JEDEC specs is 4Hi max.
Anyway a 4 controller GPU is limited to 4GB per card.

I'm wondering about the difference in the die area required between the current GDDR5 controllers and the new HBM controller, and if we can expect a future console to have 6 or 8 controller, or if 4 would be considered the limit for the expected cost target?
 
Ah, sorry I was confused. So many articles I read mixed up the capacity per die, per stack, and per system without saying which is which. And for the first time a databook expressed the chip capacity in GB instead of Gb, I thought it was per die in gb.:runaway:

But I also found this:
http://wccftech.com/amd-radeon-r9-390x-fiji-xt-8-hi-8gb-hbm/
AMD will be employing a technology by SK Hynic called a “Dual Link Interposer” which will allow memory capacities upwards of 4GB without the use of 2nd Generation HBM.
[..]
With a Dual Link Interposing design, SK Hynix will be able to stack 4x (Dual 1GB HBM modules)
[..]
2x 4-HI HBM1 (which should technically be called 8-Hi-Hi according to nomenclature rules)
So with HBM2 being 8Gbits per die, 8Hi, we could get 64GB total with 4 controllers if they do the same dual link trick?
 
If I am not wrong, Windows RT was a special breed of Windows, and not "compatible". But Windows 10 is the same for all platforms.
I just saying that for me is more posible now that there is just one ecosystem for APPs, than a year ago.
WinRT ran all "Metro apps", Win10 on ARM will run all "Universal Windows Apps" (=evolution of Metro apps more or less). It still won't run any "desktop" aka Win32 apps
 
Current console architecture is GCN much more than x86. I'd say 90% GCN and 10% x86.
And I would like it to be 99.9% GPU in the next generation.

Eh?

There's no way we won't have CPUs next gen. If anything we'll have far more powerful CPUs as a focus for next-gen, given that deminishing returns will mean a less than standard 10x jump in GPU ALU performance will be more than acceptable.
 
There's no way we won't have CPUs next gen. If anything we'll have far more powerful CPUs as a focus for next-gen, given that deminishing returns will mean a less than standard 10x jump in GPU ALU performance will be more than acceptable.
I also think we'll still have a focus on conventional CPUs but only becuase the nextegn consoles are well into planning already and right now there no evidence that sufficient work undertaking now on CPU can be deployed to GPU's compute. No doubt AMD, Intel and Nvidia have more flexible compute in mind for products shipping in 3-4 years but it'd be a hell of a gamble to commit to it now.
 
There's any idea about HSA2 features and timeframe?
Zen will be deeply integrated with the gpu compared to carrizo?
 
Status
Not open for further replies.
Back
Top